Efficient and Rapid Template-Directed Nucleic Acid Copying Using
2′-Amino-2′,3′-dideoxyribonucleoside-5′-Phosphorimidazolide
Monomers
Jason P. Schrum, Alonso Ricardo, Mathangi Krishnamurthy, J. Craig Blain, and
Jack W. Szostak*
Howard Hughes Medical Institute and Department of Molecular Biology and Center for
Computational and IntegratiVe Biology, Massachusetts General Hospital, 185 Cambridge Street,
HarVard Medical School, Boston, Massachusetts 02114
Downloaded by HARVARD UNIV on September 29, 2009 | http://pubs.acs.org
Publication Date (Web): September 16, 2009 | doi: 10.1021/ja906557v
Received August 3, 2009; E-mail: szostak@molbio.mgh.harvard.edu
Abstract: The development of a sequence-general nucleic acid copying system is an essential step in the
assembly of a synthetic protocell, an autonomously replicating spatially localized chemical system capable
of spontaneous Darwinian evolution. Previously described nonenzymatic template-copying experiments
have validated the concept of nonenzymatic replication, but have not yet achieved robust, sequence-general
polynucleotide replication. The 5′-phosphorimidazolides of the 2′-amino-2′,3′-dideoxyribonucleotides are
attractive as potential monomers for such a system because they polymerize by forming 2′f5′ linkages,
which are favored in nonenzymatic polymerization reactions using similarly activated ribonucleotides on
RNA templates. Furthermore, the 5′-activated 2′-amino nucleotides do not cyclize. We recently described
the rapid and efficient nonenzymatic copying of a DNA homopolymer template (dC15) encapsulated within
fatty acid vesicles using 2′-amino-2′,3′-dideoxyguanosine-5′-phosphorimidazolide as the activated monomer.
However, to realize a true Darwinian system, the template-copying chemistry must be able to copy most
sequences and their complements to allow for the transmission of information from generation to generation.
Here, we describe the copying of a series of nucleic acid templates using 2′-amino-2′,3′-dideoxynucleotide-5′-phosphorimidazolides. Polymerization reactions proceed rapidly to completion on short homopolymer RNA and LNA templates, which favor an A-type duplex geometry. We show that more efficiently copied
sequences are generated by replacing the adenine nucleobase with diaminopurine, and uracil with C5(1-propynyl)uracil. Finally, we explore the copying of longer, mixed-sequence RNA templates to assess
the sequence-general copying ability of 2′-amino-2′,3′-dideoxynucleoside-5′-phosphorimidazolides. Our
results are a significant step forward in the realization of a self-replicating genetic polymer compatible with
protocell template copying and suggest that N2′fP5′-phosphoramidate DNA may have the potential to
function as a self-replicating system.
Introduction
A simple synthetic cell, or protocell, must be able to grow,
divide, and maintain heritable information. We are attempting
to construct such a system through the integration of two
complementary components: (1) a spatially localized replicating
compartment or vesicle and (2) a spontaneously replicating
genetic polymer.1,2 Our group recently demonstrated that fatty
acid membranes are permeable to small-molecule solutes
including charged mononucleotides, and that activated nucleotides added to the outside of fatty acid vesicles are capable of
copying an encapsulated genetic polymer.3 We have also
demonstrated repeated cycles of growth and division for fatty
acid vesicles without loss of encapsulated material.4 However,
no general nucleic acid copying mechanism has been demonstrated with the capacity to function as a self-replicating
(1) Chen, I. A.; Roberts, R. W.; Szostak, J. W. Science 2004, 305, 1474–
1476.
(2) Szostak, J. W.; Bartel, D. P.; Luisi, P. L. Nature 2001, 409, 387–390.
(3) Mansy, S. S.; Schrum, J. P.; Krishnamurthy, M.; Tobe, S.; Treco, D. A.;
Szostak, J. W. Nature 2008, 454, 122–125.
(4) Zhu, T. F.; Szostak, J. W. J. Am. Chem. Soc. 2009, 131, 5705–5713.
10.1021/ja906557v CCC: $40.75  XXXX American Chemical Society
protocellular genome, even though many ribozyme-mediated
and nonenzymatic genetic polymer replication schemes have
been investigated.5-10 Ribozymes such as the naturally occurring
sunY and Tetrahymena self-splicing introns or the in Vitro
evolved class I and R3C ligases and polymerases are able to
assemble short oligonucleotides or mononucleotides in a
template-directed manner. All, however, lack the high reaction
efficiency and sequence-general copying ability required for the
replication of a protocell genome.11-16
(5) Xu, Y.; Kool, E. T. Nucleic Acids Res. 1999, 27, 875–881.
(6) Heuberger, B. D.; Switzer, C. Org. Lett. 2006, 8, 5809–5811.
(7) Benner, S. A.; Ricardo, A.; Carrigan, M. A. Curr. Opin. Chem. Biol.
2004, 8, 672–689.
(8) Grossmann, T. N.; Strohbach, A.; Seitz, O. Chembiochem 2008, 9,
2185–2192.
(9) Lee, D. H.; Severin, K.; Yokobayashi, Y.; Ghadiri, M. R. Nature 1997,
390, 591–594.
(10) Sievers, D.; von Kiedrowski, G. Nature 1994, 369, 221–224.
(11) Doudna, J. A.; Usman, N.; Szostak, J. W. Biochemistry 1993, 32, 2111–
2115.
(12) Muller, U. F.; Bartel, D. P. RNA 2008, 14, 552–562.
J. AM. CHEM. SOC. XXXX, xxx, 000
9
A
Downloaded by HARVARD UNIV on September 29, 2009 | http://pubs.acs.org
Publication Date (Web): September 16, 2009 | doi: 10.1021/ja906557v
ARTICLES
Schrum et al.
The alternative strategy of chemical replication was pioneered
by Orgel and colleagues who demonstrated the nonenzymatic
template-directed polymerization of 5′-phosphorimidazolideactivated ribonucleotides.17-20 This seminal work validated the
principle of enzyme-independent template copying, but also
highlighted the difficulties associated with nonenzymatic replication. In general, these chemical-copying reactions are very
slow and work best on a limited range of C-rich templates, while
A-rich template copying can only be accomplished at subfreezing temperatures.21 Furthermore, the regiospecificity of the
reaction is poor, and the products tend to have a majority of
2′f5′ linkages. Orgel initiated the use of 2′- and 3′-aminonucleotides to increase reaction rates and regiospecificity22-26
due to the enhanced nucleophilicity of the amine relative to the
normal sugar hydroxyls. Similar phosphoramidate ligation
chemistry was used to demonstrate the amplification of tetramer
and hexamer oligonucleotide templates using dimer and trimer
substrates, respectively.27,28 Recent work from Richert and
colleagues suggests that the rate of the chemical step in the
polymerization of 3′-amino nucleotides can be very fast (>1
min-1 per nucleotide) with optimal leaving group chemistry and
the use of 3′-oligonucleotides or capping groups to optimize
orientation.25,29 However, 3′-aminonucleotides are subject to
competing intramolecular cyclization.30 Moreover, the thermal
stability of a N3′fP5′-DNA duplex is even greater than that
of its RNA counterpart, suggesting that strand separation for
N3′fP5′ phosphoramidate nucleic acids may be difficult or
impossible.31,32
More divergent chemistry has been explored in the search
for an effective sequence copying system. Recent work by Liu,
Lynn and colleagues has demonstrated rapid DNA-templatedirected peptide nucleic acid (PNA) synthesis using reductive
amination to drive oligonucleotide condensation reactions.33,34
It will be of interest to see if PNA-directed PNA synthesis can
be achieved in this way. A radically different approach uses
(13) Paul, N.; Joyce, G. F. Proc. Natl. Acad. Sci. U.S.A. 2002, 99, 12733–
12740.
(14) Johnston, W. K.; Unrau, P. J.; Lawrence, M. S.; Glasner, M. E.; Bartel,
D. P. Science 2001, 292, 1319–1325.
(15) Lincoln, T. A.; Joyce, G. F. Science 2009, 323, 1229–1232.
(16) Green, R.; Szostak, J. W. Science 1992, 258, 1910–1915.
(17) Lohrmann, R.; Orgel, L. E. J. Mol. EVol. 1979, 12, 237–257.
(18) Fakhrai, H.; van Roode, J. H.; Orgel, L. E. J. Mol. EVol. 1981, 17,
295–302.
(19) Sleeper, H. L.; Lohrmann, R.; Orgel, L. E. J. Mol. EVol. 1979, 13,
203–214.
(20) Sulston, J.; Lohrmann, R.; Orgel, L. E.; Miles, H. T. Proc. Natl. Acad.
Sci. U.S.A. 1968, 59, 726–733.
(21) Vogel, S. R.; Richert, C. Chem. Commun. 2007, 1896–1898.
(22) Lohrmann, R.; Orgel, L. E. J. Mol. EVol. 1976, 7, 253–267.
(23) Zielinski, W. S.; Orgel, L. E. Nucleic Acids Res. 1985, 13, 2469–
2484.
(24) Chen, J. J.; Cai, X.; Szostak, J. W. J. Am. Chem. Soc. 2009, 131,
2119–2121.
(25) Rothlingshofer, M.; Kervio, E.; Lommel, T.; Plutowski, U.; Hochgesand, A.; Richert, C. Angew. Chem., Int. Ed. 2008, 47, 6065–6068.
(26) Lohrmann, R.; Orgel, L. E. Nature 1976, 261, 342–344.
(27) Zielinski, W. S.; Orgel, L. E. Nature 1987, 327, 346–347.
(28) von Kiedrowski, G. Angew. Chem., Int. Ed. Engl. 1986, 25, 932–
935.
(29) Rojas Stutz, J. A.; Richert, C. Chemistry 2006, 12, 2472–2481.
(30) Hill, A. R., Jr.; Nord, L. D.; Orgel, L. E.; Robins, R. K. J. Mol. EVol.
1988, 28, 170–171.
(31) Gryaznov, S. M.; Winter, H. Nucleic Acids Res. 1998, 26, 4160–4167.
(32) Mansy, S. S.; Szostak, J. W. Proc. Natl. Acad. Sci. U.S.A. 2008, 105,
13351–13355.
(33) Kleiner, R. E.; Brudno, Y.; Birnbaum, M. E.; Liu, D. R. J. Am. Chem.
Soc. 2008, 130, 4646–4659.
(34) Li, X.; Zhan, Z. Y.; Knipe, R.; Lynn, D. G. J. Am. Chem. Soc. 2002,
124, 746–747.
B J. AM. CHEM. SOC.
9
VOL. xxx, NO. xx, XXXX
reversible thioester or imine exchange to add nucleobases, in a
template-directed manner, to a preformed peptide, which
becomes the backbone of the assembled PNA-like polymer.35,36
Despite the promising initial results obtained with activated
2′-amino-ribonucleotides over 30 years ago, Orgel did not
further explore the potential of this system because the monomer
building blocks did not appear to be prebiotically plausible.
Since our immediate goal is the assembly of a self-replicating
protocell, we have decided to focus on developing an efficient
nucleic acid chemical-copying mechanism unrestrained by
prebiotic chemical considerations, and have therefore begun to
explore the potential of a series of phosphoramidate-linked
nucleic acids and their corresponding amino-nucleotide monomers in chemical-copying reactions.24,37,38 We are particularly
interested in the activated 2′-amino, 2′-3′-dideoxyribonucleotides
because previously described activated RNA mononucleotide
polymerization reactions display a bias for 2′f5′-linkages, and
because steric restrictions imposed by the ribose ring prevent
the unwanted intramolecular cyclization reaction seen with the
activated 3′-amino analogues. In addition, the 2′-amino, 2′-3′dideoxyribonucleotides have the potential to generate more
chemically stable (DNA-like) polymers than the more RNAlike 2′-amino, 3′-hydroxy nucleotides. We recently showed that
2′-amino-2′, 3′-dideoxyguanosine-5′-phosphorimidazolide (2′NH2-ImpddG) participates in the rapid and efficient templatedirected copying of a dC15 template, in a reaction that proceeds
to >95% completion in ∼6 h.3 This reaction is compatible with
the copying of templates encapsulated within fatty acid vesicles
and could thus be used in a protocell model system. However,
to enable a nucleic acid replication system that can proceed for
multiple generations, and which is capable of Darwinian
evolution, the copying reaction must be efficient and sequencegeneral, so that most if not all sequences and their complements
can be replicated. Here, we present further studies of templatedirected nucleic acid copying using 2′-amino-2′, 3′-dideoxynucleoside-5′-phosphorimidazolides (2′-NH2-ImpddNs). We
describe the copying of a series of short homopolymer templates
composed of DNA, RNA, locked nucleic acid (LNA), and
O2′fP5′ DNA and therefore varying in helical geometry, sugar
conformation, and base stacking. We describe the use of
nucleotides with modified bases to increase the base-pairing
strength and copying efficiency of the weak A:T base-pair.
Finally we examine the copying of longer mixed-sequence RNA
templates by 2′-NH2-ImpddNs.
Results and Discussion
Template-Copying Strategy. We synthesized the four standard
2′-NH2-ImpddNs (1, 3, 4, and 6) from the corresponding 2′azido nucleosides of A, U, C, and G, as described in the
Experimental Section. We then used a primer-extension assay
to test each 2′-NH2-ImpddN in a nonenzymatic templatecopying reaction (Figure 1). A 5′-Cy3-labeled 2′-aminoterminated DNA primer was annealed to a complementary
oligonucleotide with an overhanging template region. The
elongation of the fluorescent primer was used to follow the
(35) Ura, Y.; Beierle, J. M.; Leman, L. J.; Orgel, L. E.; Ghadiri, M. R.
Science 2009, 325, 73–77.
(36) Heemstra, J. M.; Liu, D. R. J. Am. Chem. Soc. 2009, 131, 11347–
11349.
(37) Hagenbuch, P.; Kervio, E.; Hochgesand, A.; Plutowski, U.; Richert,
C. Angew. Chem., Int. Ed. 2005, 44, 6588–6592.
(38) Ferencic, M.; Reddy, G.; Wu, X.; Guntha, S.; Nandy, J.; Krishnamurthy, R.; Eschenmoser, A. Chem. BiodiVersity 2004, 1, 939–979.
Downloaded by HARVARD UNIV on September 29, 2009 | http://pubs.acs.org
Publication Date (Web): September 16, 2009 | doi: 10.1021/ja906557v
Phosphoramidate DNA: Self-Replicating Genetic Polymer
ARTICLES
Figure 1. N2′fP5′-DNA genetic polymer system. (a) 2′-Amino-2′,3′-dideoxyribonucleoside-5′-phosphorimidazolides. Clockwise from upper-left: 2′-
NH2-ImpddA, R1 ) -H or 2′-NH2-ImpddD, R1 ) -NH2; 2′-NH2-ImpddC; 2′-NH2-ImpddU, R1 ) -H or 2′-NH2-Impdd-C5-(1-propynyl)-U, R1 )
-CtC-CH3; 2′-NH2-ImpddG. (b) Structure of O2′fP5′-phosphodiester-linked DNA (right) compared to N2′fP5′-phosphoramidate-DNA (left). The
2′-internucleotide bridging atom substitution is shown in red. (c) Primer-extension reaction scheme. A 5′-Cy3-labeled 2′-amino-terminated DNA primer
anneals to a complementary template. 2′-NH2-ImpddN analogues form Watson-Crick base pairs on a complementary template and participate in a chemical
chain reaction forming the N2′fP5′-DNA polymer product. (d) Chemistry of the template-directed copying reaction using the activated 2′-NH2-ImpddN
monomer. The attacking nucleophile is shown in red.
stepwise chemical addition of activated nucleotide monomers,
generating four new N2′fP5′-phosphoramidate DNA linkages.
Given the influence of template structure in template-directed
chemical-copying reactions, we tested DNA, RNA, LNA, and
O2′fP5′ DNA templates, which vary in sugar conformation
from C2′-endo in DNA, flexible C3′-endo in RNA, and O2′fP5′
DNA39 and rigid C3′-endo in LNA (the sugar in LNA is
rigidified by a methylene bridge between O2′ and C4′40,41). We
also studied the effect of Watson-Crick base-pair strength by
comparing copying reactions on templates containing adenine
to templates containing diaminopurine which form two vs three
hydrogen bonds to the complementary thymidine or uridine
nucleotide. We also compared copying reactions on uridine and
C5-(1-propynyl)uridine (Up) templates to clarify base-stacking
and template preorganization effects.42 Finally, in preparation
for future efforts to copy N2′fP5′-phosphoramidate DNA
templates, we examined primer-extension reactions on O2′fP5′phosphodiester DNA templates.
Template-Directed Activated Mononucleotide Polymerization.
We previously observed efficient template-directed copying of
a dC15 template using 2′-NH2-ImpddG (Figure 1a). This result
led us to explore structural factors that might affect the reaction
efficiency using the set of C4 nucleic acid templates listed above.
(39) Sheppard, T. L.; Breslow, R. J. Am. Chem. Soc. 1996, 118, 9810–
9811.
(40) Obika, S.; Nanbu, D.; Hari, Y.; Morio, K.-I.; In, Y.; Ishida, T.;
Imanishi, T. Tetrahedron Lett. 1997, 38, 8735–8738.
(41) Koshkin, A.; Singh, S. K.; Nielsen, P.; Rajwanshi, V. K.; Kumar, R.;
Meldgaard, M.; Olsen, C. E.; Wengel, J. Tetrahedron 1998, 54, 3607–
3630.
(42) He, J.; Seela, F. Nucleic Acids Res. 2002, 30, 5485–5496.
Primer-extension reaction products were analyzed by highresolution polyacrylamide gel electrophoresis (Figure 2). Reaction efficiency is summarized as the percent of primer extended
to full-length (g+4 nts) product at a given time point. The LNA
template, with a rigid C3′-endo sugar conformation, is the most
efficient template, reaching >98% completion within 10 min,
while the DNA template is over an order of magnitude less
efficient, taking 3 h to achieve 65% completion. Both the RNA
and the O2′fP5′-DNA templates have intermediate activity,
reaching >98% completion at 3 h. Control reactions on
noncomplementary templates showed low levels of +1 product
at 3 h, but no full-length product (Figure S3). We confirmed
the formation of the proper N2′fP5′-phosphoramidate DNA
primer-extension product on the RNA C4 template by MALDITOF MS (Figure 2). We further characterized the N2′fP5′DNA generated on the dC15 template (Figure S1) by showing
that it exhibited the expected phosphoramidate bond acid-lability
(Figure S1), but is resistant to enzymatic digestion by RQ1
DNase and RNase If (Figure S2).
Given the efficient template copying observed with the
activated guanosine amino nucleotide, we next assessed the
activity of 2′-amino-2′,3′-dideoxycytidine-5′-phosphorimidazolide (2′-NH2-ImpddC) (Figure 1a) on complementary DNA,
RNA, LNA, and O2′fP5′-DNA templates with the sequence
G4. We observed very efficient copying on the RNA, LNA, and
O2′fP5′-DNA templates; in all cases the reactions were
essentially complete within 10 min (Figure 3). We observed
no extension on the DNA template, possibly due to aggregation
of the dG4 region to form G-quartets. Control extension reactions
on noncomplementary templates show limited extension to the
J. AM. CHEM. SOC.
9
VOL. xxx, NO. xx, XXXX
C
Downloaded by HARVARD UNIV on September 29, 2009 | http://pubs.acs.org
Publication Date (Web): September 16, 2009 | doi: 10.1021/ja906557v
ARTICLES
Schrum et al.
Figure 2. Primer-extension reaction using a 2′-NH2-ImpddG as monomer. (a) Primer-extension reaction scheme showing a 5′-Cy3-labeled 2′-aminoterminated DNA primer annealed to a complementary template. 2′-NH2-ImpddG monomer participates in a chemical chain reaction extending the primer
by four (4) nucleotides on the complementary template forming a N2′fP5′-DNA polymer product. * indicates new phosphoramidate bond. (b) Highresolution gel electrophoresis analysis of primer-extension products on indicated templates. Primer-extension reactions contained 0.1 µM Cy3-labeled-2′amino-terminated DNA primer, 0.5 µM template, 100 mM MES-CAPS-HEPES, pH 7.5, 100 mM 1-(2-hydroxyethyl)imidazole, and 5 mM 2′-NH2-ImpddG.
The reaction was initiated by addition of 2′-NH2-ImpddG, thermal cycling as described in the Experimental Section, and incubation at 4 °C for the indicated
time. Arrows indicate primer and full-length product. (c) MALDI-TOF MS analysis of primer-extension product. 100-200 pmol amino-terminated primer
was extended on an RNA template followed by base hydrolysis of the RNA template and preparation for MALDI-TOF MS as detailed in the Experimental
Section. DNA control oligonucleotide (left): calcd mass 4999.87 and observed mass 4999.67; N2′fP5′-DNA extension product (right): calcd mass 4994.95
and observed mass 4994.91. (d) Graphical summary of primer-extension reactions. Primer-extension reaction efficiency is expressed as percent full-length
product (g+4) achieved at the given time point.
+1 position with traces of +2 product but do not generate any
full-length +4 product (Figure S4, Supporting Information).
MALDI-TOF MS analysis confirmed the expected N2′fP5′DNA primer-extension product (Figure 3).
We then asked if 2′-amino-2′,3′-dideoxyadenosine-5′-phosphorimidazolide (2′-NH2-ImpddA) (Figure 1a) would efficiently copy DNA T4, RNA U4, LNA T4, and O2′fP5′-DNA
T4 templates. These primer-extension reactions were very slow
and inefficient compared to our previous results with G and C.
Increasing the concentration of 2′-NH2-ImpddA from 5 mM
to 30 mM made very little difference (Figure 4). We observed
significant accumulation of the full-length +4 product only on
the LNA template at 24 h. Because thymidine and uridine have
less effective base-stacking interactions than purine nucleotides,
we prepared templates containing four successive residues of
the deoxyuridine analogue C5-(1-propynyl)deoxyuridine (dUp)
in an attempt to preorganize the template geometry through
stabilizing π-stacking interactions between 5-propynyl groups
in adjacent nucleotides.42 However, the dUp4 template only
modestly improves reaction efficiency when compared to the
dT4 template (Figure 4).
We then tested template-copying reactions based on the
reciprocal A:T/U base pair by synthesizing 2′-amino-2′,3′dideoxyuridine-5′-phosphorimidazolide (2′-NH2-ImpddU) (Figure 1a) and examining primer-extension reactions on DNA A4,
RNA A4, LNA A4, and O2′fP5′-DNA A4 templates. As above,
all reactions were of low efficiency at both 5 mM and 30 mM
2′-NH2-ImpddU (Figure 5). While slow, both RNA and LNA
D J. AM. CHEM. SOC.
9
VOL. xxx, NO. xx, XXXX
templates did lead to the accumulation of some full-length +4
product at 24 h (Figure 5). In an effort to improve reaction
efficiency, we sought to strengthen the A:T/U base-pair by
preparing templates in which adenine was replaced with 2,
6-diaminopurine (D), which contains an exocyclic amine at the
C-2 position and has the potential to make three hydrogen bonds
with the complementary Watson-Crick base, thymidine or
uridine. However, the DNA-D4 template did not enhance
reaction efficiency significantly when compared with the dA4
template (Figure 5).
Improved Template Copying with D:Up Base-Pair. Since the
2′-NH2 A and U nucleotides performed so poorly in templatedirected copying reactions, and since the C5-(1-propynyl)-U and
diaminopurine nucleobase modifications had such modest effects
when present only in the template, we prepared the activated
monomers for D (2) and Up (5) and tested these modified
nucleotides on templates containing D and Up. The D:Up basepair has previously been reported to be as effective as a G:C
base-pair in the context of the nonenzymatic synthesis of RNA
on a DNA template.43 We observed a wide range of reaction
efficiencies with the 2′-NH2-ImpddD monomer (Figure 1a) on
DNA T4, DNA dUp4, RNA U4, RNA Up4, LNA T4 and
O2′fP5′-DNA T4 templates (Figure 6). The improvement
expected from Up containing templates was much more evident
when the monomer was 2′-NH2-ImpddD. The dUp4 template
(43) Chaput, J. C.; Sinha, S.; Switzer, C. Chem. Commun. 2002, 1568–
1569.
Downloaded by HARVARD UNIV on September 29, 2009 | http://pubs.acs.org
Publication Date (Web): September 16, 2009 | doi: 10.1021/ja906557v
Phosphoramidate DNA: Self-Replicating Genetic Polymer
ARTICLES
Figure 3. Primer-extension reaction using a 2′-NH2-ImpddC as monomer. (a) Primer-extension reaction scheme. * indicates new phosphoramidate bond.
(b) High-resolution gel electrophoresis analysis of primer-extension products on indicated templates. Primer-extension reactions were completed as previously
described with 5 mM 2′-NH2-ImpddC. Arrows indicate primer and full-length product. (b) MALDI-TOF MS analysis of primer-extension product performed
as described in the Experimental Section. DNA control oligonucleotide (left): calcd mass 4839.84 and observed mass 4839.66; N2′fP5′-DNA extension
product (right): calcd mass 4834.92 and observed mass 4835.08. (c) Graphical summary of primer-extension reaction efficiency.
Figure 4. Primer-extension reaction using a 2′-NH2-ImpddA as monomer. (a) Primer-extension reaction scheme. * indicates new phosphoramidate bond.
(b) High-resolution gel electrophoresis analysis of primer-extension products on indicated templates. Template-copying reactions were run as previously
described with 5 mM (upper) or 30 mM (lower) 2′-NH2-ImpddA. Arrows indicate primer and full-length extension product.
achieves 15% full-length product at 3 h compared to 0% full-length
product on the standard T4 DNA template. A similar enhancement
is seen on the RNA templates, where the RNA Up4 template reaches
82% full-length product compared 40% full-length product on the
standard RNA-U4 template in 3 h. Interestingly, the RNA U4 and
O2′fP5′-DNA T4 templates display similar reaction efficiencies,
with >80% full-length product accumulation over the course of
24 h, again demonstrating that 2′f5′-linked oligonucleotides serve
as effective templates. Experiments at 5 and 30 mM 2′NH2-ImpddD gave similar results (data not shown) suggesting
that the template is already saturated with 5 mM monomer.
Noncomplementary template copying reactions yield only small
amounts of +1 product and no full-length product at 3 h (Figure
S5, Supporting Information). We detected the proper N2′fP5′DNA primer-extension polymerization product by MALDI-TOF
MS analysis (Figure 6).
J. AM. CHEM. SOC.
9
VOL. xxx, NO. xx, XXXX
E
ARTICLES
Schrum et al.
Downloaded by HARVARD UNIV on September 29, 2009 | http://pubs.acs.org
Publication Date (Web): September 16, 2009 | doi: 10.1021/ja906557v
Figure 5. Primer-extension reaction using the 2′-NH2-ImpddU as monomer. (a) Primer-extension reaction scheme. * indicates new phosphoramidate bond.
(b) High-resolution gel electrophoresis analysis of primer-extension products on indicated templates. Primer-extension reactions were run as described with
5 mM (upper) or 30 mM (lower) 2′-NH2-ImpddU. Arrows indicate primer and full-length product.
Figure 6. Primer-extension reaction using the 2′-NH2-ImpddD as monomer. (a) Primer-extension reaction scheme. * indicates new phosphoramidate bond.
(b) High-resolution gel electrophoresis analysis of primer-extension products on indicated templates. Primer-extension reactions were completed as previously
described with 5 mM 2′-NH2-ImpddD. (c) MALDI-TOF MS analysis of primer-extension product. DNA control oligonucleotide (left): calcd mass 4935.89
and observed mass 4936.00; N2′fP5′-DNA extension product (right): calcd mass 4991.01 and observed mass 4990.86. (d) Graphical summary of primerextension reactions.
We prepared the second unnatural nucleobase activated
analogue, 2′-amino-2′,3′-dideoxy-C5-(1-propynyl)uridine-5′phosphorimidazole (2′-NH2-ImpddUp) and used it in primerextension experiments to copy a series of A and D containing
templates: DNA A4, DNA D4, RNA A4, RNA D4, LNA A4,
and O2′fP5′-DNA A4 templates (Figure 7). Copying of the
LNA and RNA A4 templates was largely complete after 3 and
24 h respectively. The effect of enhanced hydrogen bonding
on reaction efficiency was clearly visible by comparing the RNA
A4 template to the RNA D4 template. The RNA D4 template
reaction is largely complete in 10-20 min compared to 24 h
for the RNA A4 template. Control copying reactions on
F
J. AM. CHEM. SOC.
9
VOL. xxx, NO. xx, XXXX
noncomplementary templates result in traces of +1 product but
no full-length product (Figure S6, Supporting Information). We
confirmed synthesis of the proper primer-extension product on
the RNA D4 template by MALDI- TOF MS (Figure 7).
Mixed-Sequence RNA Template-Copying Reactions. Because
our goal is to develop a sequence-general self-replicating nucleic
acid system, we examined the ability of 2′-NH2-ImpddN
monomers to copy longer mixed-sequence templates. We used
RNA templates for these experiments because of their synthetic
accessibility; in future work we plan to examine N2′fP5′-DNA
templates. We synthesized RNA templates containing the natural
guanosine and cytidine nucleobases and the unnatural diami-
Downloaded by HARVARD UNIV on September 29, 2009 | http://pubs.acs.org
Publication Date (Web): September 16, 2009 | doi: 10.1021/ja906557v
Phosphoramidate DNA: Self-Replicating Genetic Polymer
ARTICLES
Figure 7. Primer-extension reaction using a 2′-NH2-ImpddUp as monomer. (a) Primer-extension reaction scheme. * indicates new phosphoramidate bond.
(b) High-resolution gel electrophoresis analysis of primer-extension products on indicated templates. Primer-extension reactions were completed as previously
described with 5 mM 2′-NH2-ImpddUp. (c) MALDI-TOF MS analysis of primer-extension product. DNA control oligonucleotide (left): calcd mass 4899.84
and observed mass 4899.84; N2′fP5′-DNA extension product (right): calcd mass 4990.92 and observed mass 4990.94. (d) Graphical summary of primerextension reactions.
nopurine and C5-(1-propynyl)uridine nucleobases. We first
examined RNA templates containing two noncomplementary
nucleotides: 5′-GGGUpUpUp- and 5′-CCCDDD- downstream of
the primer binding site (Figures S7 and S8, Supporting Information). On both templates, primer-extension reactions led to the
appearance of full-length product (g+6 nts) within 60 min and
were largely complete by 3 h. Because N2′fP5′-DNA cannot
be sequenced using standard sequencing techniques that require
transcription into DNA or RNA, we used single-nucleotide
dropout primer-extension reaction experiments to begin to assess
the fidelity of the template copying reaction. On both of these
templates, the primer-extension reaction stops at the expected
location in the absence of one of the 2′-NH2-ImpddN monomers, with only small amounts of primer extended by a single
mismatched base present at 3 h.
We then proceeded to examine two more complex RNA
templates containing three different nucleotides in the templating
region. Both sequences (5′-GGDDCCDDCCCDDGG- and 5′AGCCGCCGCC-3′- adjacent to the primer binding region)
allowed full-length product accumulation, but the primerextension reaction was significantly slower than the previously
examined templates. Only a trace amount of full-length product
is present at 1 h, while the reactions required over 20 h to
accumulate significant full-length product (Figures S9 and S10,
Supporting Information). Individual 2′-NH2-ImpddN dropout
reactions indicate the potential for significant problems with the
fidelity of template copying. In the absence of the G monomer,
primer-extension stops cleanly before the first C residue on both
templates. However, in the absence of the C monomer, there is
extensive read-through of G residues on the 5′-GGDDCCDDCCCDDGG-template (Figure S9, Supporting Information),
presumably due to wobble-pairing of template Gs with Up
monomers but also possibly due to the formation of G:G
mismatches. In the absence of the Up monomer, ∼50% of the
primer stalled before the first template D residue, as expected,
but a similar amount is stalled after the incorporation of a
mismatched base (either G or C, or both) opposite the first
template D (Figure S9, Supporting Information). On the template
sequence 5′-CCGCCGCCGC-, which was copied with just the
ImpddG and ImpddC monomers, we see extensive primer
extension in the absence of the C monomer, which must be
due to the incorporation of G opposite G (Figure S10, Supporting Information). These G:G mismatches appear to be more
slowly extended than G:C base-pairs, resulting in the accumulation of multiple stalled primer-extension products with altered
gel mobility (Figure S10, Supporting Information), presumably
because the G:G pair distorts the helix and prevents the proper
alignment of the 2′-amino group of the terminal G residue for
attack on the 5′-phosphorimidazolide of the incoming activated
nucleotide.
Finally, we asked if a mixed-sequence RNA template
containing all four nucleotides could be copied. We used a
template with the sequence 5′-CDDCCDGUpCDDCDCG-3′
adjacent to the primer binding site to model a sequence-generalcopying reaction (Figure 8). We used a mixture of the
2′-NH2-ImpddG, 2′-NH2-ImpddC, 2′-NH2-ImpddD, and 2′NH2-ImpddUp monomers at optimized concentration ratios.
These ratios were derived from a series of primer-extension
experiments carried out at various concentration ratios, in which
we attempted to maximize full-length product accumulation and
minimize stalling. Using these optimized monomer ratios, we
were able to observe conversion of 61% of primer to full-length
primer-extension product through nonenzymatic templatedirected synthesis of N2′fP5′-DNA. Full-length (g+15 nts)
product is visible at the 3 h time point and continues to
accumulate up to 24 h (Figure 8).
J. AM. CHEM. SOC.
9
VOL. xxx, NO. xx, XXXX
G
ARTICLES
Schrum et al.
that copying of C and D is likely to proceed with relatively
high fidelity.
Downloaded by HARVARD UNIV on September 29, 2009 | http://pubs.acs.org
Publication Date (Web): September 16, 2009 | doi: 10.1021/ja906557v
Conclusions
Figure 8. High-resolution gel electrophoresis analysis of mixed-sequence
RNA template-copying reaction. (a) Primer-extension reaction scheme. A
5′-Cy3-labeled 2′-amino-terminated DNA primer annealed to a complementary RNA template. 2′-NH2-ImpddN monomers participate in a
chemical chain reaction extending the primer by 15 nucleotides on the
complementary template, forming a N2′fP5′-DNA polymer product. *
indicates new phosphoramidate bond. (b) Primer-extension reactions were
completed as previously described with 0.5 µM template (5′-CDDCCDGUpCDDCDCG-3′-primer binding site), 5 mM 2′-NH2-ImpddG, 5 mM
2′-NH2-ImpddC, 10 mM 2′-NH2-ImpddD, and 10 mM 2′-NH2-ImpddUp.
The reaction was initiated with addition of the monomers, thermal cycling
as described in the Experimental Section, and incubation at 4 °C for the
indicated time. Arrows indicate primer, full-length product, and stalled
products.
We used single-nucleotide dropout experiments as above to
assess potential problems with fidelity (Figure 8). In the absence
of the D or C monomers, full-length product is still formed,
presumably due to the G:Up wobble base-pairing. In the absence
of D, a +7 nucleotide primer-extension product accumulates,
suggesting that copying of the single Up in the template by G
monomer is slow compared to copying by the correct D
monomer. In the absence of the C monomer, full-length primerextension product accumulates to similar levels as in the
complete reaction mix. In this case, a stalled +1 product is
visible at 24 h but is decreased in intensity by 72 h. This suggests
that the Up monomer may incorporate opposite the first G
residue in the template, but that this wobble base-pair is further
extended relatively slowly. Very little stalled product is visible
either before or at the position of the second G residue in the
template. Thus, the incorporation of U opposite template G
residues may be a significant source of errors. In contrast, in
the absence of G or Up, very strong stalling is observed prior
to the first template C and D residue, respectively, suggesting
H J. AM. CHEM. SOC.
9
VOL. xxx, NO. xx, XXXX
Our exploration of the spontaneous chemical copying of a
range of nucleic acid templates by activated 2′-amino 2′,3′dideoxyribonucleotide monomers highlights both the advantages
and potential problems associated with this approach to template
copying. Here we discuss the implications of copying experiments using templates of varying length and sequences,
composed of varying nucleic acids. On the basis of these results
we suggest directions for future work that may lead to improved
sequence-general copying and ultimately an effective selfreplication system.
In our initial set of experiments, we synthesized the 2′-amino2′,3′-dideoxy analogues of the four standard ribonucleotides,
A, G, C, and U, and activated each of these by forming the
corresponding 5′-phosphorimidazolides. We used each of these
compounds to extend a primer across a complementary fournucleotide template region using DNA, RNA, LNA, and
O2′fP5′-DNA templates. We observed that optimal templatecopying efficiency is a strong function of template sugar
conformation and preorganization. In all cases, the order of
template activity was LNA > RNA ∼ O2′fP5′-DNA > DNA,
strongly suggesting that a preorganized A-type helical structure
provides the optimal context for template-directed polymerization. These initial experiments also showed that G4 and C4
templates were very rapidly copied by activated C and G
monomers, respectively. On LNA and RNA templates, these
primer-extension reactions were complete in less than 10 min,
which is at least 100-fold faster than comparable primer
extension reactions on RNA templates using identically activated
ribonucleotides.44 Thus, we see the expected rate enhancement
due to the replacement of the normal hydroxyl nucleophile of
ribonucleotides with a 2′-amino nucleophile. This rate enhancement is similar to that reported earlier for 3′-amino nucleotides.25
When we examined A4 and U4 (or T4) templates, we found
that these were copied very inefficiently by activated 2′-amino
U and A monomers; only LNA and to a lesser extent RNA
templates supported full-length primer-extension, and even in
these cases the full-length product accumulated very slowly.
The A:T(U) base-pair is well-known to be less stabilizing to
duplex formation for two reasons: the presence of only two
hydrogen bonds between A and T(U) and the weaker stacking
interactions of U (and T) compared to the other nucleotides. We
attempted to correct these deficiencies by replacing A with
diaminopurine (D), which allows for the formation of a base-pair
with three hydrogen bonds, and by replacing U(T) with C5-(1propynyl)-U, where the propynyl substituent enhances stacking
interactions. Either modification alone had little effect, but together
the results were more dramatic; the copying of the RNA D4
template by the 2′-amino Up monomer was complete in 10-20
min. However, the copying of the RNA Up template was much
slower and was not complete until 3 h. Although the Up monomer
is clearly superior to U or T, a further improved derivative would
clearly be useful. The electron-withdrawing propynyl substituent
has previously been noted to reduce the pKa of N3 by nearly one
unit relative to the N3 of uridine, thus reducing the strength of an
A:Up base-pair relative to A:T in another system.45 A less electronwithdrawing substituent at the 5 position, that could still enhance
(44) Ninio, J.; Orgel, L. E. J. Mol. EVol. 1978, 12, 91–99.
Downloaded by HARVARD UNIV on September 29, 2009 | http://pubs.acs.org
Publication Date (Web): September 16, 2009 | doi: 10.1021/ja906557v
Phosphoramidate DNA: Self-Replicating Genetic Polymer
the stacking properties of the nucleobase, might yield improved
template-copying performance.
We noticed a significant reduction in the rate of primerextension on our mixed-sequence RNA templates. For example,
the G, C, and D 4-base homopolymer templates were all copied
in ∼10 min, but the 6-base mixed templates took 1-3 h and
the 15- and 16-base templates were not completely copied after
24 h. There are several possible explanations for this effect.
Template secondary structures or template-template interactions
may prevent the activated 2′-NH2-ImpddNs from aligning on
the template, although the sequences we used should support
only weak interactions. If this explanation is correct, decreased
template concentration and/or elevated temperature during the
primer-extension reaction may facilitate the copying reaction.
Alternatively, binding of 2′-NH2-ImpddNs to incorrect bases
on the template may decrease the polymerization rate, either
by preventing primer-extension when a mismatched base is
bound, or because the incorporation of mismatched bases into
the growing chain slows subsequent chain elongation. It may
be possible to test this hypothesis using specially designed
templates and varying monomer concentrations and ratios.
Another possibility is that the slow copying of longer templates
may be due to the weak pairing of the newly synthesized
N2′fP5′-DNA region with the RNA template, so that the
terminus of the growing chain becomes increasingly likely to
dissociate from the template as the chain grows longer. This
hypothesis could be tested with homopolymer templates of
varying lengths. Finally, it is possible that the regular helical
geometry of homopolymer templates is disrupted with mixed
sequences, perhaps due to irregular stacking interactions of
template bases or bound monomers. Modified nucleobases with
improved stacking interactions might be necessary to overcome
such an effect.
Successful genetic inheritance requires high fidelity replication. Early work by Eigen and more recent work by Nowak
and Ohtsuki has explored the fidelity required for successful
template replication.46,47 In essence, successful transmission of
genetic information can only occur if the sequence mutation
rate, u, is less than the inverse of the length of the sequence
(1/n), no matter how great the selective advantage conferred
by that sequence. Copying a functional + strand into a
complementary - strand, then back to a new + strand, requires
two rounds of strand copying, and thus the per nucleotide, per
strand error rate should be less than 1/2n. This requirement may
be alleviated by a factor of approximately 2 when considering
realistic functional sequences, such as ribozymes, where not
every base need be strictly defined to retain function.48 Thus,
replication of functional sequences of 30 to 35 nucleotides would
seem to require an error rate of less than about 3%. As noted
above, the most likely source of errors during template copying
with 2′-amino nucleotides is G:Up wobble pairing. Primerextension past template G residues in the absence of C monomer
suggests that G:Up wobble pairing may compete effectively with
G:C Watson-Crick pairing. However, it is not possible from
our current results to estimate the error frequency resulting from
wobble-pair-mediated misincorporation, since this depends on
the competition between correct and incorrect monomer in
(45) Chen, J. J.; Tsai, C. H.; Cai, X.; Horhota, A. T.; McLaughlin, L. W.;
Szostak, J. W. PLoS ONE 2009, 4, e4949.
(46) Eigen, M.; Schuster, P. Naturwissenschaften 1977, 64, 541–565.
(47) Nowak, M. A.; Ohtsuki, H. Proc. Natl. Acad. Sci. U.S.A. 2008, 105,
14924–14927.
(48) Kun, A.; Santos, M.; Szathmary, E. Nat. Genet. 2005, 37, 1008–1011.
ARTICLES
reactions where both are present. Accurate measurement of
copying fidelity will require the ability to examine the products
of primer-extension by techniques that are more informative
then gel electrophoresis. It is unfortunate that the standard
methods of molecular biology such as cloning and sequencing
are not applicable in this case, since N2′fP5′-DNA cannot at
present be accurately copied into DNA by any known DNA
polymerase. However, high-resolution MS/MS methods should
allow the composition of N2′fP5′-DNA oligonucleotides to
be accurately determined. We note that there are at least two
possibilities for a reduction in errors due to G:Up wobble pairing.
First, an improved version of C5-(1-propynyl)-U with stronger
base-pairing to D would allow the use of lower concentrations
of the U monomer, thus decreasing G:Up wobble pairing. A
second, more speculative possibility, is that fidelity may be
higher on N2′fP5′-DNA templates. The geometry of the
N2′fP5′-DNA duplex, which is still unknown, may be such
that wobble pairing greatly disfavors the chemical step in chain
elongation. We hope to be able to investigate this possibility in
the near future.
On the basis of the results discussed above, we believe that
N2′fP5′-DNA is an attractive candidate for development into
a self-replicating genetic system. Most significantly, the activated
2′-NH2-ImpddN monomers copy a variety of nucleic acid
templates on the scale of minutes in efficient primer-extension
reactions. Furthermore, since the monomers do not cyclize, the
major competing reaction is hydrolysis, which is quite slow (kobsd
<10-2 h-1 at 37 °C, pH 7.5).49 This allows ample time for the
completion of template-copying chemistry before the substrates
are depleted by hydrolysis. Finally, in the context of multiple
replication cycles, where duplex strand separation must occur
before the next round of template copying can begin, the
N2′fP5′-DNA duplex is likely to melt at a lower temperature
than the corresponding N3′fP5′-DNA duplex.50 Oligonucleotides containing 12 to 16 O2′fP5′ linkages have Tm values
between 10 and 30 °C lower than that of O3′fP5′ linked
sequences, depending on the G-C content. Facile strand separation, and possible slower strand reannealing, could greatly
facilitate the emergence of cycles of chemical replication. Our
current work is focused on the synthesis of N2′fP5′-DNA
oligonucleotides so that we can explore the possibility of true
self-replication in this system.
Regardless of whether or not the N2′fP5′-DNA system
ultimately proves capable of robust self-replication, the work
presented here provides guidance for the design of alternative
and potentially superior systems. Most relevant is the striking
superiority in template-copying reactions of the conformationally
constrained, and therefore preorganized, LNA templates. This
suggests that other conformationally constrained templates may
show enhanced chemical replication properties. N2′fP3′-TNA
has been synthesized by the Eschenmoser laboratory and is
known to be a Watson-Crick base-pairing system.38 It is
conformationally constrained relative to N2′fP5′-DNA because
there is one less rotatable bond per backbone repeat unit. A
second alternative is the morpholino nucleic acid system, which
is conformationally constrained in a very different way, due to
the presence of a six-membered ring in the backbone repeat
unit. We suggest that these and other conformationally con(49) Kanavarioti, A.; Bernasconi, C. F.; Doodokyan, D. L.; Alberas, D. J.
J. Am. Chem. Soc. 1989, 111, 7247–7257.
(50) Hashimoto, H.; Switzer, C. J. Am. Chem. Soc. 1992, 114, 6255–6256.
J. AM. CHEM. SOC.
9
VOL. xxx, NO. xx, XXXX
I
ARTICLES
Schrum et al.
strained nucleic acids are excellent candidates for chemically
self-replicating genetic systems.
Downloaded by HARVARD UNIV on September 29, 2009 | http://pubs.acs.org
Publication Date (Web): September 16, 2009 | doi: 10.1021/ja906557v
Experimental Section
General Methods. Analytical thin layer chromatography (TLC)
was performed on Merck silica gel 60 matrix plates containing a
fluorescent indicator. Compounds were characterized by proton (1H),
carbon (13C), and phosphorus (31P) nuclear magnetic resonance
(NMR) spectroscopy on a Varian NMR spectrometer (Oxford AS400) (1H: 400 MHz, 13C: 100 MHz, and 31P: 80 MHz). In each
case the chemical shift of the corresponding solvent was used as a
reference. Chemical shift values are reported in ppm. Electrospray
mass spectrometry was recorded on a Bruker Daltonics Esquire
6000 ESI-MS. Chemicals and solvents were purchased from SigmaAldrich unless otherwise indicated.
Crude 2′-amino-2′,3′-dideoxynucleoside-5′-phosphorimidazolides were purified by preparative HPLC (Varian ProStar
Preparative LC) on a Prep-C18 column 21.2 mm × 150 mm, 5 µ
particle size (Agilent Technologies) equilibrated with 25 mM
triethylammonium bicarbonate/2.5% acetonitrile, pH 8.0 and eluted
with an acetonitrile gradient followed by further purification by
reverse-phase HPLC (Agilent 1100 series LC) on a Zorbax ExtendC18 column 9.4 mm × 250 mm, 5 µ particle size (Agilent
Technologies) equilibrated with 30 mM triethylammonium bicarbonate/2% acetonitrile, pH 8.0, and eluted with an acetonitrile
gradient. Purified imidazolides were lyophilized to dryness and
stored at -80 °C.
Oligonucleotides were synthesized using standard β-cyanoethyl
phosphoramidite chemistry on an Expedite 8909 DNA synthesizer
(Applied Biosystems) unless otherwise indicated. β-cyanoethyl
phosphoramidites used in oligonucleotide synthesis were purchased
from Glen Research unless otherwise noted. DNA synthesizer
reagents were purchased from Glen Research and American
Bioanalytical. Oligonucleotide quantification was done on a Nanodrop ND1000 Spectrophotometer.
Abbreviations. Pd/C, palladium on carbon; DMF, dimethylformamide; TOM, [(triisopropylsilyl)oxy]methyl; lcaa-CPG, long-chain
alkylamine-Controlled Pore Glass bead; MES, 2-(N-morpholino)ethanesulfonic acid; CAPS, N-cyclohexyl-3-aminopropanesulfonic
acid; HEPES, 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid.
2′-Amino-2′,3′-dideoxy-adenosine-5′-Phosphorimidazole (2′NH2-ImpddA) (1). 2′-Azido-2′,3′-dideoxyadenosine was synthesized according to the reported protocol.51 2′-Azido-2′,3′-dideoxyadenosine (276.1 mg, 1 mmol) was dissolved in freshly distilled
triethylphosphate (10 mL) and cooled to 0 °C in an ice bath under
an argon atmosphere. Phosphorus oxychloride (1.2 equiv) was added
dropwise and the solution stirred at 0 °C until complete consumption
of the nucleoside, as determined by TLC. Imidazole (4 equiv) was
then added, the solution was stirred at 0 °C for 30 min and then
allowed to warm to room temperature and stirred for an additional
3 h. The volatiles were evaporated, and DMF (1 mL) was added to
dissolve the residue. The 2′-azido adenosine phosphorimidazolide
was precipitated by dropwise addition into a solution of diethyl
ether/acetone/triethylamine (17 mL, 2:1.2:0.15) containing NaClO4
(25 mg) and collected by centrifugation. The crude 2′-azido
adenosine phosphorimidazolide was dissolved in methanol (10 mL)
containing triethylamine (1 mL), and the azido function was reduced
by catalytic hydrogenation on Pd/C (10% Pd loading, 0.1 mmol)
over a period of 2 h. The suspension was filtered over Celite, the
volatiles were evaporated, and the product was purified by
preparative RP-HPLC, yielding 193.7-220.9 mg (70-80%) of 1
as a white solid. 1H NMR δ (D2O): 8.01 (1H, s), 7.52 (1H, s), 6.82
(1H, s), 6.65 (1H, s), 5.70 (1H, s), 4.44 (1H, m), 3.91 (1H, m),
3.84 (1H, m), 3.68 (1H, m), 2.15 (1H, m), 1.91 (1H, m). 31P δ
(D2O): -6.80. ESI-MS: mass calcd: (-) 379.10; found: (-) 379.1.
2′-Amino-2′,3′-dideoxy-2,6-diaminopurine-5′-Phosphorimidazole (2′-NH2-ImpddD) (2). 2′-Azido-2′,3′-dideoxy-2,6-diaminopurine
was prepared according to the reported protocol.51 2 was prepared
according to the protocol for compound 1. 1H NMR δ (D 2O): 7.52
J
J. AM. CHEM. SOC.
9
VOL. xxx, NO. xx, XXXX
(1H, s), 7.35 (1H, s), 6.61 (1H, s), 6.47 (1H, s), 4.27 (1H, m), 3.76
(1H, m), 3.65 (1H, m), 3.52 (1H, m), 2.03 (1H, m), 1.75 (1H, m). 31P
δ (D2O): -6.89. ESI-MS: mass calcd: (-) 394.11; found: (-) 394.1.
2′-Amino-2′,3′-dideoxycytidine-5′-Phosphorimidazole (2′-NH2ImpddC) (3). 2′-Azido-2′,3′-dideoxycytidine was prepared according to the reported protocol.51 3 was prepared according to the
protocol for compound 1 with the following exceptions. The crude
2′-azido-2′,3′-dideoxycytidine-5′-phosphorimidazolide was dissolved in a solution of pyridine/triethylamine/water (10 mL, 85:
13:2) and cooled to 0-5 °C in an ice bath. The solution was then
saturated with hydrogen sulfide gas, and the resulting brown mixture
was stirred for 30 min. The volatiles were removed by evaporation
under reduced pressure (CAUTION: vacuum pump must be inside
fume hood. Hydrogen sulfide is extremely toxic. Low concentrations
have a strong odor, but higher levels ablate the sense of smell and
lead to loss of consciousness followed by death) and the residue
coevaporated with anhydrous toluene (2×, 10 mL). The crude
product was then purified by RP-HPLC. 1H NMR δ (D2O): 7.72(1H,
s), 7.49 (1H, d, J ) 7.6 Hz), 6.91 (1H, s), 5.78 (1H, d, J ) 8 Hz),
5.52 (1H, br s), 4.35 (1H, m), 3.96 (1H, m), 3.72 (1H, m), 3.35
(1H, m), 1.85 (1H, m), 1.69 (1H, m). 31P δ (D2O): -6.83. ESIMS: mass calcd: (-) 355.09; found: (-) 355.1.
2′-Amino-2′,3′-dideoxyuridine-5′-Phosphorimidazole (2′-NH2ImpddU) (4). 2′- Azido-2′,3′-dideoxyuridine was prepared according
to the reported protocol.51 4 was prepared according to the protocol
for compound 1. 1H NMR δ (D2O): 7.72 (1H, s), 7.38 (1H, d, J )
7.6 Hz), 7.09 (1H, s), 6.91 (1H, s), 5.58 (1H, d, J)7.8 Hz), 5.54
(1H, d, 4 Hz), 4.30 (1H, m), 3.91 (1H, m), 3.71 (1H, m), 3.34 (m,
1H), 1.87 (m, 1H), 1.71 (m, 1H). 31P δ (D2O): -6.85. ESI-MS:
mass calcd: (-) 356.08; found: (-) 356.1.
2′-Amino-2′,3′-dideoxy-5-C-propynyluridine-5′-Phosphorimidazole (2′-NH2-ImpddUp) (5). 2′-Trifluoroacetamide-2′,3′-dideoxyC5-propynyluridine synthesis details will be described elsewhere. 2′Trifluoroacetamide-2′,3′-dideoxy-C5-propynyluridine (361.1 mg, 1
mmol) and Proton-Sponge (2 equiv) were dissolved in freshly distilled
triethylphosphate (10 mL) and cooled to 0 °C in an ice bath under an
argon atmosphere. Phosphorus oxychloride (1.2 equiv) was then added
dropwise and the solution stirred at 0 °C until complete consumption
of the nucleoside, as determined by TLC. Imidazole (4 equiv) was
then added, the solution was stirred at 0 °C for 30 min and then allowed
to warm up to room temperature and stirred for additional 3 h. The
volatiles were then evaporated, and DMF (1 mL) was added to the
residue. The phosphorimidazolide was precipitated by dropwise
addition to a solution of diethyl ether/acetone/triethylamine (17 mL,
2:1.2:0.15) containing NaClO4 (25 mg) and collected by centrifugation.
The crude product was dissolved in methanol (10 mL) containing
K2CO3 (200 mM) and stirred overnight at room temperature. After
deprotection, the desired product was purified by RP-HPLC as
previously described.1H NMR δ (D2O): 7.57 (1H, s), 7.54 (1H, s),
6.96 (1H, s), 6.72 (1H, s), 5.35 (1H, d, J ) 2.8 Hz), 4.19 (1H, m),
3.75 (1H, m), 3.61 (1H, m), 3.26 (1H, m), 1.76 (1H, m), 1.67 (3H, s),
1.59 (1H, m). 31P δ (D2O): -7.0. ESI-MS: mass calcd: (-) 394.09;
found: (-) 394.1.
2′-Amino-2′,3′-dideoxyguanosine-5′-Phosphorimidazole (2′NH2-ImpddG) (6). 2′-Azido-2′,3′-dideoxyguanosine was synthesized
according to the reported protocol.51 6 was prepared according to the
protocol for compound 1. 1H NMR δ (D2O): 7.47 (1H, s), 7.37 (1H,
s), 6.66 (1H, s), 6.49 (1H, s), 5.32 (1H, d, J ) 3.2 Hz), 4.21 (1H, m),
3.71-3.62 (2H, m), 3.50 (1H, m), 1.98 (1H, m), 1.70 (1H, m). 31P
(D2O) -6.87. ESI-MS: mass calcd: (-) 395.1; found: (-) 395.1.
5′-O-Dimethoxytrityl-5-propyn-1-yl-uridine-2′-O-Triisopropylsilyloxymethyl-3′-[(2-cyanoethyl)-(N,N-diisopropyl)]phosphoramidite (7). 5′-O-DMT-2′-O-TOM-5-C-propynyluridine nucleoside was
synthesized in four steps as previously described.52 5′-O-DMT-2′O-TOM-5-C-propynyluridine nucleoside (953 mg, 1.23 mmol) was
(51) Kawana, M.; Kuzuhara, H. J. Chem. Soc. Perkin Trans. 1 1992, 4,
469–478.
Downloaded by HARVARD UNIV on September 29, 2009 | http://pubs.acs.org
Publication Date (Web): September 16, 2009 | doi: 10.1021/ja906557v
Phosphoramidate DNA: Self-Replicating Genetic Polymer
dissolved in anhydrous dichloromethane (20 mL) and cooled to
0-5 °C in an ice bath under an argon atmosphere. Diisopropylethylamine (0.64 mL, 3.7 mmol) and cyanoethyl diisopropylphosphoramidochloridite (0.43 g, 1.85 mmol) were then added and the
resulting solution stirred at room temperature. After 2 h, TLC
analysis showed complete conversion of the starting material into
a more polar compound, and the reaction mixture was diluted with
additional dichloromethane (60 mL) and extracted with aqueous
NaHCO3 (30 mL) and washed with water (30 mL). The organic
phase was dried with MgSO4, and after filtration the volatiles were
removed under reduced pressure. The residue was purified by
column chromatography using a gradient of 10-60% ethyl acetate
in hexane. The product was obtained as a pair of diastereomers.
1
H NMR δ (CDCl3): 7.98 and 7.92 (2 s, 1H), 7.15-7.48 (m, 10
H), 6.77-6.87 (m, 4H), 6.16 and 6.11 (2 d, 1H, J ) 6.8 Hz and J
) 6.0 Hz), 4.92-5.02 (2 m, 2H), 4.51 and 4.57 (2 m, 2H), 4.44
and 4.38 (2 m, 1H), 4.27 and 4.17 (2 m, 1H), 3.80-3.96 (2 m,
1H), 3.77 and 3.76 (2 s, 6H), 3.32-3.7 (m, 6H), 2.60-2.66 (m,
1H), 2.3-2.4 (m, 1H), 1.65-1.68 (m, 3H), 1.11-1.18 (m, 12H),
0.96-1.06 (m, 18 H).13C NMR δ (CDCl3): 162.16, 162.07, 149.69,
149.59, 144.73, 144.53, 142.06, 141.79, 135.87, 135.81, 135.56,
130.35, 130.25, 130.21, 130.17, 128.27, 128.24, 128.15, 128.11,
127.15, 127.13, 113.52, 113.49, 101.77, 101.58, 91.26, 91.09, 87.60,
87.53, 87.33, 84.34, 83.23, 70.56, 69.98, 63.19, 55.47, 55.45, 24.87,
24.85, 24.81, 24.77, 18.05, 17.97, 12.15, 12.10, 12.07, 4.68, 4.66.
31
P NMR δ (CDCl3): 151.77, 151.24. ESI-MS (-): 969.5; (+):
971.3.
Oligonucleotide Synthesis. DNA and LNA template oligonucleotides (5′-NxGCAGTCAGTCTACGC-3′) were synthesized using
standard β-cyanoethyl phosphoramidite chemistry. RNA oligonucleotides were either purchased from Dharmacon, Inc. or
synthesized in-house using TOM-protected RNA phosphoramidites.
DNA and LNA oligonucleotides were cleaved from the synthesis
column and base-deprotected in 40% aqueous ammonium hydroxide
at 55 °C for 17 h. dUp DNA templates were synthesized with dmfdG phosphoramidite to reduce ammonium hydroxide deprotection
time, thereby avoiding cyclization of the dUp residues. Integrity of
the dUp templates was confirmed by MALDI-TOF MS and lack of
fluorescence. RNA oligonucleotides were cleaved from the synthesis
column and base-deprotected in 1:1 aqueous methylamine:ethanolic
methylamine (40% methylamine) solution for 10 min at 65 °C. RNA
oligonucleotide 2′-TOM protecting groups were removed using 1
M tetrabutylammonium fluoride in tetrahydrofuran for at least 6 h
at 35 °C. DNA and LNA oligonucleotides were purified by anionexchange HPLC on a DNAPac PA-200 column (Dionex) in 10 mM
Tris-Cl, pH 8.0 in a gradient up to 2 M KCl. RNA oligonucleotides
were purified by denaturing polyacrylamide gel electrophoresis. The
amino-terminated DNA primer (5′-Cy3-GCGTAGACTGACTGC2′-amino) was synthesized as previously described.3 5′-H-phosphonate-2′-deoxythymidine-3′-lcaa-CPG (total loading, 20 µmol)
was charged with 5′-dimethoxytrityl-2′-amino-2′,3′-dideoxycytidine51 (100 µmol, 5 equiv) via oxidative amination in 1:1 carbon
tetrachloride/acetonitrile. The primer synthesis proceeded by standard β-cyanoethyl phosphoramidite chemistry with a 5′-terminal
Cy3-phosphoramidite dye added last. Primer was cleaved from the
column and base-deprotected in 40% aqueous ammonium hydroxide
at room temperature for at least 36 h. The primer was then treated
with 80% acetic acid for 16 h at room temperature to hydrolyze
the 2′-phosphoramidate linkage and generate the 2′-amino-5′-Cy3labeled DNA primer. The primer was purified by anion-exchange
HPLC on the DNAPac PA-200 column in 10 mM NaOH, pH 12,
against a 1.5 M NaCl gradient. All oligonucleotides were subsequently purified on an Oligonucleotide Purification Cartridge
(Applied Biosystems) to remove excess salt from DNA and LNA
(52) Berry, D. A.; Jung, K.-Y.; Wise, D. S.; Sercel, A. D.; Pearson, W. H.;
Mackie, H.; Randolph, J. B.; Somers, R. L. Tetrahedron Lett. 2004,
45, 2457–2461.
ARTICLES
oligonucleotides and traces of acrylamide, urea, and salt from the
RNA oligonucleotides.
Primer-Extension Reactions. Template-copying reactions contained 0.1 µM Cy3-labeled 2′-amino-terminated primer, 0.5 µM
template oligonucleotide, 150 mM NaCl, 100 mM 1-(2-hydroxyethyl)imidazole, 100 mM Na+-MES-CAPS-HEPES, pH 7.5 and 2′-amino2′,3′-dideoxynucleoside-5′-phosphorimidazolide at indicated concentrations. Reactions were initiated by addition of monomer followed
by thermal cycling at 55 °C for 30 sf4 °C for 2 min repeated twice
followed by incubation at 4 °C for the duration of the reaction time.
Primer extension reactions incubated at 4 °C without thermal cycling
left a significant amount of nonutilized 2′-amino-terminated DNA
primer. Thermal cycling yielded full primer utilization presumably
due to rearrangement of primer/template complexes trapped in
nonproductive conformations. Nonreactive DNA primer/template
complexes may require a shift of the 2′-amino terminus into a
C3′-endo conformation, a possibility confirmed in pilot primer extension reactions using a 2′-amino-terminated RNA primer where all
the sugars are in the C3′-endo conformation, which does not require
thermal cycling for full utilization (data not shown). Aliquots were
removed and stopped at indicated time points by adding three
volumes formamide and heating to 95 °C for 2 min followed
immediately by ethanol precipitation on dry ice. Stopped reactions
were resuspended in formamide gel loading buffer and heated to 95
°C for 5 min. Samples were analyzed by electrophoresis on 7 M
urea, 17% polyacrylamide sequencing gels. Reaction products were
visualized by fluorescence imaging on a Typhoon 9410 PhosphorImager using the Cy3 fluorophore filter set. Product quantification and
analysis was performed using ImageQuant TL software (GE Healthcare Life Sciences).
Primer Extension Product Mass Spectrometry. Primer-extension product mass was determined by extending 100-200 pmol of
the amino-terminated DNA primer on a complementary RNA
template in 150 mM NaCl, 100 mM 1-(2-hydroxyethyl)-imidazole,
100 mM Na+-MES-CAPS-HEPES, pH 7.5 and 5 mM 2′-amino2′,3′-dideoxynucleoside-5′-phosphorimidazolide at 4 °C. The
completed primer-extension reaction was subjected to base hydrolysis in 600 mM KOH at 50 °C for 2 h to degrade the RNA
template leaving the DNA primer with four 2′-terminal 2′-5′phosphoramidate linkages and a new 2′-amino terminus. The KOH
hydrolysis reaction mixture was then ethanol precipitated and
resuspended in water in preparation for MALDI-MS. The sample
was desalted and washed with 0.1 M triethylamine acetate on a
C18 ZipTip (Millipore). MALDI matrices were 50 mg/mL 3-hydroxypicolinic acid or 10 mg/mL 2,4,6-hydroxyacetophenone with
0.1 M diammonium citrate in 50% acetonitrile. Samples were
spotted directly on the MALDI-TOF stainless steel plate and the
mass acquisition was completed in the positive mode on a MALDITOF mass spectrometer (PerSeptive Biosystems, Model Voyager
DE).
Acknowledgment. We are grateful to Jesse J. Chen, Yollete
V. Guillen-Schlippe, Ben Heuberger, Sylvia Tobé, Doug Treco,
Tina Yuan, and Shenglong Zhang for helpful discussions. A.R. is
a Research Associate and J.W.S. is an Investigator of the Howard
Hughes Medical Institute. This work was supported in part by NSF
Grant CHE 0434507 to J.W.S.
Supporting Information Available: Phosphoramidate acid and
enzymatic sensitivity analysis and experimental section (Figures
S1 and S2); primer extension reactions on noncomplementary
templates (Figures S3-S6); additional mixed-sequence RNA
template-copying reactions (Figures S7-S10). This material is
available free of charge via the Internet at http://pubs.acs.org.
JA906557V
J. AM. CHEM. SOC.
9
VOL. xxx, NO. xx, XXXX
K